Jon's Blog

Logo

Blog for thoughts and ideas, when a Tweet isn't enough.

View My GitHub Profile

15 April 2018

Why hyperconvergence worked for us

by jon

Well, two and a half reasons. The first being the semi-unique situation that we found ourselves in. The trouble began, not a few years ago, but much farther back. Big purchases, meant to solve all the problems, and therein lies the first rub. Over subscription, taking what you have and making a best guess if it will fit tomorrows requirements. This always fails, it is very challenging to predict the future, its not quite folly to do so, but take your lumps.

Reason #1 (the half reason is in here too)

We had a string of big purchase that we never completely migrated off of. One big array, two years later a set of arrays meant to mirror each other (under Novell!), then later another big array, without the first being fully retired. In total, a decent amount of storage, but in chunks, which makes it difficult to provide a large amount of contiguous storage. Running in parallel to this were the switches, again, starting with 4, then adding 4 more, only retiring two of the older. Never completing the migration off the old. Add in understaffing and you have a recipe for a unfinished upgrades over multiple generations of hardware.

Reason #2

Very little to actually change. The solution we chose (Nutanix), didn’t rely on an previous investments. It was a completely clean break from storage and compute. It allowed us to stage up a discrete environment, using existing VMware licenses and perform migration to the new systems. Since it was using VMware, there wasn’t a need to learn a new hypervisor. Since the Prism management interface made provisioning a cluster a 10 minute process (That we got assistance with on install), the migration was pretty close to painless. Any additional expansions that came after were equally painless and quick.

Thoughts

Removing complexity is a big bonus of hyperconvergence. I’m able to add more to my clusters without any thoughts to storage oriented upgrades (arrays and san switches). Since they are flexible in nature, unless I look at a complete forklift upgrade, I’m only aging out portions of the over all system.

back

tags: infrastructure people